Claude can now read samples of your team's conversations and produce a report

On October 28, 2026, Anthropic made smart reports generally available for Claude Enterprise, the business version of the Claude chatbot. Two other announcements came at the same time: in-app surveys and shared group budgets.

According to Anthropic, Claude reads a sample of a team's conversations within a selected scope and writes a report on what the work consisted of. The report sorts the work into areas and shows usage, deliverables and friction points. It also gives recommendations for recurring workflows and rules for code repositories. Users can ask their own questions and save them as templates, and the findings can be shared.

Deloitte employee Vivek Kulkarni is quoted in Anthropic's announcement: "Smart reports give us greater insight into how AI is used across the company."

What can the boss see?

Anthropic's help article describes a number of built-in safeguards. Conversations about restricted topics are used only in aggregate, with no summaries of individual conversations and no names. This applies to personal matters, health, union activity, harassment, whistleblowing and political views, among other things. Custom questions may not be about these topics.

The inefficiency section counts sessions that produced nothing useful, as total numbers and costs. Individual sessions cannot be viewed.

The organization chooses whether viewers see usernames or simply "User." If names are turned on, it is logged when they are revealed. Anthropic writes that the reports are not designed for, and should not be used for, evaluating individuals' performance or making hiring decisions.

Who can use it?

Only a Primary Owner or Owner can enable the feature. A Primary Owner, Owner, Admin or a custom role with access to analytics can create and view reports. Delegates can create reports for assigned groups, departments or cost centers, and according to Anthropic, team leads who are not admins can run reports for their groups.

Each organization can create 10 reports per month at no charge. The feature is not available for organizations with customer-managed encryption keys (CMEK) or HIPAA configuration, among others.

According to the independent review AI Codex, the feature launched as a beta on September 10, 2026, and a report can cover up to 28 days. AI Codex recommends limiting reports to specific teams and running the same scope twice, four weeks apart, to measure change.

Surveys and shared budgets

The in-app surveys ask standard questions about value, time savings and operational significance. Admins control recipients and timing, and responses can be exported to spreadsheets.

With shared monthly group budgets, heavy users avoid hitting their limit while others leave budget unused. Individual limits remain, and requests for more budget go to admins. Email alerts are sent at 50, 75, 95 and 100 percent of the budget.